TracingPaper.tex

% template-v1.tex: LaTeX2e template for Usenix papers.
% Version: usetex-v1, 31-Oct-2002
% Revision history at end.

\documentclass[XXX,endnotes]{usetex-v1}
% Choose the appropriate option:
%
% 1. workingdraft:
%
%       For initial submission and shepherding.  Features prominent
%       date, notice of draft status, page numbers, and annotation
%       facilities.  The three supported annotation macros are:
%               \edannote{text}         -- anonymous annotation note
%               \begin{ednote}{who}     -- annotation note attributed
%                 text                          to ``who''
%               \end{ednote}
%               \HERE                   -- a marker that can be left
%                                               in the text and easily
%                                               searched for later
% 2. proof:
%
%         A galley proof identical to the final copy except for page
%         numbering and proof date on the bottom.  Annotations are
%         removed.
%
% 3. webversion:
%
%       A web-publishable version, uses \docstatus{} to indicate
%       publication information (where and when paper was published),
%       and page numbers.
%
% 4. finalversion:
%
%       The final camera-ready-copy (CRC) version of the paper.
%       Published in conference proceedings.  This doesn't include
%       page numbers, annotations, or draft status (Usenix adds
%       headers, footers, and page numbers onto the CRC).
%
% If several are used, the last one in this list wins
%

%
% In addition, the option "endnotes" permits the use of the
% otherwise-disabled, Usenix-deprecated footnote{} command in
% documents.  In this case, be sure to include a
% \makeendnotes command at the end of your document or
% the endnotes will not actually appear.
%

% These packages are optional, but useful
\usepackage{epsfig}     % postscript figures
\usepackage{url}        % \url{} command with good linebreaks

\begin{document}

\title{Tracing SMB: Searching for Unknowns}

% document status: submitted to foo, published in bar, etc.
\docstatus{Submitted to Cool Stuff Conference 2002}

% authors.  separate groupings with \and.
\author{
\authname{Paul A.\ Wortman}
\authaddr{ECE}
\authaddr{University of Connecticut}
\authaddr{ Storrs, CT, 06279}
\authurl{\url{paul.wortman@engr.uconn.edu}}
\authurl{\url{http://host.dom/yoururl}}
%\and
%\authname{Name Two}
%\authaddr{Two's Institution}
%\authurl{\url{two@host.dom}}
%
} % end author

\maketitle

\begin{abstract}
With any sort of benchmark, there are inherent oversimplifications that are taken into account when first designing these watermarks for advancing technology.  In the case of networking benchmarks, many of these simplifications occur when dealing with the low level operations of the system; spatial/temporal scaling, timestamping, IO and system behavior.  While these simplifications were acceptable for past systems being tested, this facile outlook is no longer acceptable for supplying worthwhile information.  Without taking into account the intricacies of current day machines, technology will only be able to progress in the avenues that we know of, while never being able to tackle the bottlenecks that are made apparent through more accurate benchmarking.
\end{abstract}

\section{Introduction}
\label{Introduction}
Benchmarks are important for the purpose of developing and taking accurate metrics of current technologies.  Benchmarks allow for the stress testing of different aspects of a system (e.g. network, single system).  There are three steps to creating a benchmark; first one takes a trace of an existing system.  This information is then used to compare the expected actions of a system (theory) against the traced actions of said system (practice).  The next step is to determine which aspects of the trace are most representative of what occurred during the tracing of the system, while figuring out which are represntative of the habits and patterns of said system.  This discovered information is used to produce a benchmark, either by running a repeat of the captured traces or by using synthetic benchmark created from the trends detailed within the captured tracing data~\cite{Anderson2004}.

[ADD THIS SECTION TO RELATED WORK?]As seen in previous trace work done [Leund et al, ellard et al, roselli et al], the general perceptions of how computer systems are being used versus their initial purpose have allowed for great strides in eliminating actual bottlenecks rather than spending unnecessary time working on imagined bottlenecks.  Leung's \textit{et. al.} work led to a series of obervations, from the fact that files are rarely re-opened to finding that read-write access patterns are more frequent ~\cite{Leung2008}.  Without illumination of these underlying actions (e.g. read-write ratios, file death rates, file access rates) these issues can not be readily tackled.

The purpose of my work is to tackle this gap and hopefully bring insight to the complexity of network communication.  I/O benchmarking, the process of comparing I/O systems by subjecting them to known workloads, is a widespread pratice in the storage industry and serves as the basis for purchasing decisions, performance tuning studies, and marketing campaigns ~\cite{Anderson2004}.

\subsection{Issues with Tracing}
\label{Issues with Tracing}
The majority of benchmarks are attempts to represent a known system and structure on which some “original” design/system was tested.  While this is all well and good, there are many issues with this sort of approach; temporal \& spatial scaling concerns, timestamping and buffer copying, as well as driver operation for capturing packets~\cite{Orosz2013,Dabir2008,Skopko2012}.  Each of these aspects contribute to the inital problems with dissection and analysis of the captured information.  Inaccuracies in scheduling I/Os may result in as much as a factor of 3.5 differences in measured response time and factor of 26 in measured queue sizes; differences that are too large to ignore~\cite{Anderson2004}.  [MENTION EXAMPLE ISSUES BROUGHT FROM THIS - TWO GOOD EXAMPLES].

With the matter of temporal scaling, the main concern is that current day benchmarks do not account for the subtleties of intercommunication between clients \& servers on a network.  Temporal scaling refers to the need to account for the nuances of timing with respect to the run time of commands; consiting of computation, communication \& service.  A temporally scalable benchmarking system would take these subtleties into account when expanding its operation across multiple machines in a network.  While these temporal issues have been tackled for a single processor (and even somewhat for cases of multi-processor), these same timing issues are not properly handles when dealing with inter-network communication.  Spatial scaling refers to the need to account for the nuances of expanding a benchmark to incorporate a number of (\textbf{n}) machines over a network.  A system that properly incorporates spatial scaling is one that would be able to inccorporate communication (even in varying intensities) between all the machines on a system, thus stress testing all communicative actions and aspects (e.g. resource ocks, queueing) on the network.  Common practice is to have this singular benchmark run in parallel across some N computer systems \& to take the result as a facile representation of a parallel/networks system; thus the more interesting data (e.g. inter-network communication) is not accurately represented and nothing can be done about inter-network bottlenecks because these issues are not even known.

[CLOSING SENTENCES?]While performing a benchmark on a single machine is easily feasible, there is much more to consider when dealing with multiple machines communicating with each other, and the expected requirements of fully testing these aspects

\subsection{Previous Advances Due to Testing}
\label{Previous Advances Due to Testing}
Previous tracing work has shown that one of the largest \& broadest hurdles to tackle is that benchmarks must be tailored (to every extent) to the system being tested.  There are always some generalizations taken into account but these generalizations can also be a major source of error~\cite{Anderson2004,Traeger2008,Vogels1999,Dabir2008,Orosz2013,Skopko2012,Ellard2003,EllardLedlie2003,Ruemmler1993}.  To produce a benchmark with high fidelity one needs to understand not only the technology being used but how it is being implemented within the system to benchmark~\cite{Roselli2000,Ruemmler1993,Traeger2008}.  All of these aspects will lend to the behavior of the system; from timing \& resource elements to how the managing software governs~\cite{Ellard2003,EllardLedlie2003,Douceur1999}.  Further more, in persuing this work one may find unexpected results and learn new things through examination~\cite{Leung2008,Ellard2003,Roselli2000}.

[PERHAPS USE THIS PART?]Understanding that no paper an really see the whole scope of tracing/benchmarks, this paper attempts to tackle an aspect of trying to bridge macro and micro benchmarks by building a system that incorporates a micro benchmark's low level replication fidelity with proper scaling to allow for macro level and a "full spectrum scope" analysis of everything in-between using traces of data input and synthetic trace generation.  Due to the magnitude of this goal, this paper will further limit its focus towards the often forgot [CITE NEEDED?] networking aspect of multi-system scalable benchmarking and tracing.

\subsection{The Need for a New Study}
\label{The Need for a New Study}
As has been pointed out by past work, the design of systems is usually guided by an understanding of the file system workloads and user behavior~\cite{Leung2008}.  It is for that reason that new studies are constantly performed by the science community, from large scale studies to individual protocol studies~\cite{Leung2008,Ellard2003,Anderson2004,Roselli2000,Vogels1999}.  Even within these studies, the information gleaned is only as meaningful as the considerations of how the data is handled.  The following are issues that our work hopes to alleviate: there has been no large scale study done on networks for some time, there has been no study on CIFS(Common Internet File System)/SMB(Server Message Block) protocols for even longer, and most importantly these studies have not tackled lower level aspects of the trace, such as spacial \& temporal scaling idiosyncrasies of network communication.  It is for these reasons that we have developed this tracing system and have developed new studies for lower level aspects of communication network.  A detailed overview of the tracings and analysis system can be seen in section ~\ref{Tracing System}.  The hope is to further the progress made with benchmarks \& tracing in the hope that it too will lend to improbvng and deepening the knowledge and understanding of these systems so that as a result the technology and methodology is bettered as a whole.

\section{Methodology}
\label{Methodology}

\subsection{System Limitations}
\label{System Limitations}
When initially designing the tracing system used in this paper, different aspects were taken into account, such as space limitations of the tracing system, packet capture limitations (e.g. file size), and speed limitations of the hardware.  The major space limitation that is dealt with in this work is the amount of space that the system has for storing the captured packets, including the resulting DataSeries-file compressions.  One limitation encountered in the packet capture system deals with the functional pcap (packet capture file) size.  The concern being that the pcap files only need to be held until they have been filtered for specific protocol information and then compressed using the DataSeries format, but still allow for room for the DataSeries files being created to be stored.  Other limitation concerns came from the software and packages used to collect the network traffic data~\cite{Orosz2013,Dabir2008,Skopko2012}.  These ranged from timestamp resolution provided by the tracing system's kernel~\cite{Orosz2013} to how the packet capturing drivers and programs (such as dumpcap and tshark) operate along with how many copies are performed and how often.   These aspects were tackled by installing PF\_RING, which is a kernel module which allows for kernel-based capture and sampling with the idea that this will limit packets loss and timestamp overhead leading to faster packet capture while efficiently preserving CPU cycles~\cite{PFRING}.  The speed limitations of the hardware are dictated by the hardware being used (e.g. GB capture interface) and the software that makes use of this hardware (e.g. PF\_RING).  After all, our data can only be as accurate as the information being captured~\cite{Ellard2003,Anderson2004}.
Other concerns deal with the whether or not the system would be able to function optimally during periods of high network traffic.  All apsects of the system, from the hardware to the software, have been altered to help combat these concerns and allow for the most accurate packet capturing possible.

\subsection{Main Challenges}
\label{Main Challenges}
Challenges include: Interpretation of data, selective importance of information, arbitrary distribution of collected information.
One glaring challenge with building this tracing system was using code written by others; tshark \& DataSeries.  While these programs are used within the tracing structure (which will be further examined in section ~\ref{Tracing System}) there are some issues when working with them.  These issues ranged from data type limitations of the code to hash value \& checksum miscalculations due to encryption of specific fields/data.  Attempt was made to dig and correct these issues, but they were so inherrent to the code being worked with that hacks and workaround were developed to minimize their effect.  Other challenges centralize around selection, intrepretations and distribution scope of the data collected.  Which fields should be filtered out from the original packet capture?  What data is most prophetic to the form and function of the network being traced?  What should be the scope, with respect to time, of the data being examined?  Where will the most interesting information appear?  As each obstacle was tackled, new information and ways of examining the data reveal themselves and with each development different alterations \& corrections are made.

\subsection{Interpretation of Data}
\label{Interpretation of Data}
Unfortunately benchmarks require that the person(s) creating the benchmark determines the interpretation of the data collected.  To some degree these interpretations are easy to make (e.g. file system behavior \& user behavior~\cite{Leung2008}) while others are more complicated (e.g. temporal scaling of occurances of read/write), but in all scenarios there is still the requirment for human interpretation of the data.  While having humans do the interpretations can be adventageous, a lack of all the "background" information can also lead to incorrectly interpreting the information.  The hope of this project is that, despite the possible pitfall of incorrect data interpretation, we will be able to not only find out more about the workings and uses of a network but also produce a meaningful benchmark that will more accurately represent the low level aspects of large communication networks.

\subsection{Scope of Interpretation}
\label{Scope of Interpretation}
Expanding on the previous point about interpretation of data, another human factor of benchmark creation is selecting which information is important or which information will give the greatest insight to the workings on the network.  As stated earlier too little information can lead to incorrect conclusions being drawn about the workings on the system, while too much information (and not knowing which information is pertinent) can lead to erroneous conclusions as well.  Thus there is a need to strike a balance between what information is important enough to capture (so as not to slow down the capturing process through needless processing) while still obtaining enough information to acquire the bigger picture of what is going on.  Unfortunately every step of the tracing process requires a degree of human input to decide what network information will end up providing the most complete picture of the network communication and how to interpret that data into meaningful graphs and tables.  This can lead to either finds around the focus of the work being done, or even lead to discoveries of other phenomena that end up having far more impact on the overall performance of the system~\cite{Ellard2003}.

Even when all the information is collected and the most important data has been selected, there is still the issue of what lens should be used to view this information.  Because the data being collected is from an active network, there will be differing activity depending on the time of day, week, and scholastic year.  For example,  although the first week or so of the year may contain a lot of traffic, this does not mean that trends of that period of time will occur for every week of the year (except perhaps the final week of the semester).  The trends and habits of the network will change based on the time of year, time of day, and even depend on the exam schedule.  For these reasons one will see different trends depending on the distribution of the data used for analysis, and the truly interesting examination of data requires looking at all different periods of time to see how all these factors play into the communications of the network.

\section{Tracing System}
\label{Tracing System}

\subsection{Stages of Trace}
\label{Stages of Trace}

\subsubsection{Capture}
\label{Capture}
The packet capturing aspect of the tracing system is fairly straight forward.  On top of the previously mentioned alterations to the system (e.g. PF\_RING), the capture of packets is done through the use of \textit{tshark}, \textit{pcap2ds}, and \textit{inotify} programs.  The broad strokes are that incoming SMB/CIFS information comes from the university's network.  All packet and transaction information is passed through a duplicating switch that then allows for the tracing system to capture these packet transactions over a 10 Gb port.  The reason for using 10Gb hardware is to help ensure that the system is able to capture and \& all information on the network.  These packets are then passed along to the \textit{tshark} packet collection program (which is the terminal version of wireshark) which records these packets into a cyclical capturing ring.  A watchdog program (called \textit{inotify}) watches the directory where all of these packet-capture (pcap) files are being stored and as a new pcap file is completed \textit{inotify} passes the file to \textit{pcap2ds} along with what protocol is being examined (i.e. SMB).  The \textit{pcap2ds} program reads through the given pcap files, filters out any data fields deemed important or interesting for the passed protocol type, then the results are written in DataSeries format and these compressed files are then collected and stored.  Due to the fundamental nature of this work, there is no need to track every piece of information that is exchanged, only that information which illuminates the behavior of the clients \& servers that function over the network (e.g. read \& write transactions).  It should also be noted that all sensitive information being captured by the tracing system in encrypted to proect the users whose information is be examined by this tracing system.

\subsubsection{Collection}
\label{Collection}
The collection of these files is rather straight forward.  Once the DataSeries files have been collected to an arbitrary amount (in this case 100,000 files), these files are then moved off of the tracing system and are stored on a more secure \textit{/trace-store/} machine.  This storage location is only accessable from the trace system along with RAIDing on its disk to protect against data loss of the collected DataSeries files.  These files are then used in analysis to determine the behavior on the university network.

\subsubsection{Dissection/Analysis}
\label{Dissection/Analysis}
The trace analysis in performed by an analysis module code that both processes the DataSeries files for extraction of information but also outputs meaningful information (such as IO patterns) to a file that can be used for further analysis.  This section of the tracing system is always growing and changing as discoveries and limitations are found during the continuous execution of this code.  Alterations range from edits to speed up the analysis process to adjustments to how communications are tracked and interpreted.  This analysis will eventually incorporate oplocks and other aspects of resource sharing on the network to gain a more complete picture of the network's usage and bottlenecks.

\section{Trace Analysis}
\label{Trace Analysis}
The trace analysis is performed by an AnalysisModule code that both processes the ds-files for extraction of information to an event\_data structure and also outputs meaningful information (such as the IAT times between request and response packets) to a file that can be used for further analysis.

\subsection{SMB}
\label{SMB}
Server Message Block (SMB) is the modern dialect of Common Internet File System (CIFS).  The most important aspect of SAMBA (e.g. SMB) is that it is a stateful protocol , i.e. one where the information being sent via SMB has identifying fields that allow for process ID tracking.
\\The structure for sending message payloads in SMB is as follows: each SMB message is split into three blocks.  The first block is a fixed-length SMB header.  The second block is made up of two variable-length blocks called the SMB parameters.  The third block is made up of the SMB data.  Depending on the transaction occurring these different blocks are used in different manners.  The purpose of the SMB header is particularly important because the header identifies the message as an SMB message payload~\cite{MS-CIFS}.  When used in a response message the header also includes status information that indicates whether and how the command succeeded or failed.  The most important aspects of the SMB header, which the tracing system constantly examines, are the PID/MID tuple (for the purpose of identifying a client/server) and the commands value which is passed (notifying our tracing system of the actions taking place on the network).  It is through this command field that the process ID tracking system is able to follow the different commands (read/write/general event) that occur and try to find patterns in these network communications.
\\\textit{\textbf{Expectations}}: SMB will be heavily used by students to access their network accounts from any networked computer, along with network access to shared file systems and connected printers.  Oplocks will be in heavy use and cause a slowdown of the system for multiuser shared storage space.  Authentication of network computers could bottleneck during moments of high traffic (e.g. students all logging in for a class).

\subsection{ID Tracking}
\label{ID Tracking}
All comands sent over the network are coupled to an identifying MID/PID/TID/UID tuple.  Since the only commands being examined are read or write commands, the identifying characteristic distinguishing a request command packet from a reponse command packet is the addition of an FID field with the sent packet.  It is examination of the packets for this FID field that allows the analysis code to distinguish between request \& response command pakets. The pairing is done by examining the identifying tuple and assuming that each tuple-identified system will only send one command at a time (awaiting a response before sending the next command of that same type).
\\Following these process IDs is as a way to check for intercommunication between two or more processes.  In particular, we examine the compute time \& I/O (input/output) time (i.e. time spent in communication; between information arrivals).  This is done by examining the inter-arrival times (IAT) between the server \& the client. This is interesting because this information will give us a realistic sense of the data transit time of the network connections being used (e.g. ethernet, firewire, fibre, etc.).  Other pertinent information would be how often the client makes requests \& how often this event occurs per client process ID, identifiable by their PID/MID tuple.  One could also track the amount of sharing that is occurring between users.  The PID is the process identifier and the MID is the multiplex identifier, which is set by the client and is to be used for identifying groups of commands belonging to the same logical thread of operation on the client node.
\\The per client process ID can be used to map the activity of given programs, thus allowing for finer granularity in the produced benchmark (e.g. control down to process types ran by individual client levels).  Other features of interest are the time between an open \& close, or how many opens/closes occurred in a window (e.g. a period of time).  This information could be used as a gauge of current day trends in filesystem usage \& its consequent taxation on the surrounding network.  It would also allow for greater insight on the r/w habits of users on a network along with a rough comparison between other registered events that occur on the network.  Lastly, though no less important, it would allow us to look at how many occurrences there are of shared files between different users, though one must note that there is some issue (though hopefully rare) of resource locking (e.g. shared files) that needs to be taken into account.  This is initially addressed by monitoring any oplock flags that are sent for read \& writes.  This information also helps provide a preliminary mapping of how the network is used and what sort of traffic populates the communication.

%\subsection{Other (e.g. HTML)}
%\label{Other (e.g. HTML)}
%
%\subsection{Process ID Tracking}
%\label{Process ID Tracking}
%
%\begin{figure}[htbp]
%\begin{centering}
%\epsfig{file=communications_sketch, width=2.50in}
%\small\itshape
%\caption{\small\itshape Rough Sketch of Communication}
%\label{fig-communication}
%\end{centering}
%\end{figure}
%
%Figure~\ref{fig-communication} shows a rough sketch of communication between a client \& server.  The general order that constitutes a full tracking is as follows: (client) computation [process to filesystem], (client) communication [SMB protocol used to send data client→server], (server) timestamping + service [server gets data, logs it, performs service], (server) communication [SMB data send server→client], (client) next computation.  
%
%Currently the focus of process ID tracking is to see the number of reads, writes and events that occur due to the actions of clients on the network.  This is done by using a tuple of the PID \& MID fields which allows for the identification of client.  Since these values are unique and \textbf{MUST} be sent with each packet, this tuple is used as the key for the unordered map that is used to track this information.  The structure is as follows: the tuple functions as the key for the pairing of the identifying tuple \& corresponding event\_data structure; which is used to house pertinent information about reads/writes/events.  The information stored in the structure is the last time a read/write/event occurred, the total IAT of the observed read/write/events, and the total number of reads/writes/events that have occurred for the identified tuple.  The purpose for tracking this information is to profile the read/write “habits” of the users on the network as well as comparing this information against the general events’ inter-arrival times, thus allowing one to see if the read \& write events are being processed differently (e.g. longer or shorter IATs) than the rest of the events occurring on the network.
%
%One should note that there are separate purposes to the PID/MID tuple from the PID/MID/TID/UID tuple.  The first tuple (2-tuple) is used to uniquely identify groups of commands belonging to the same logical thread of operation on the client node, while the latter tuple (4-tuple) allows for unique identification for request \& responses that are part of the same transaction.  While the PID/MID tuple is mainly what we are interested in, since this allows the following of a single logical thread, there is some interest in making use of the TID/UID tuple because this would allow us to count the number of transactions that occur in a single logical thread.  This information could provide interesting information on how the computer systems on the network may be deciding to handle/send commands over the network; e.g. sending multiple commands per transaction, multiple packet commands per transaction, etc.
%
%\subsubsection{event\_data Structure Tracking}
%\label{event_data Structure Tracking}
%The purpose of the event\_data structure is to maintain a list of the interesting information associated with each PID/MID/TID/UID tuple seen on the network.  It is through this structure that the read \& write times, IATs, and even number of occurances are tracked, along with the request/response IAT pairings.  In this manner each tuple has the following information tracked, and both the packet processing is performed and the meaningful data is output from the AnalysisModule code.  \textit{\textbf{ADD LIST OF event\_data INFORMATION HERE}}.  Although there is a large number of aspects that can be examined when dealing with all of this network information, the current focus of this paper is to examine the possible read/write commands that can occur in via SMB protcols and the IAT times of the request and response packets for these commands. \textit{\textbf{Note:}} Eventually the addition of resource locks WILL be included because it is through this information that we can gain any sort of idea as to the interaction between users/other programs with the resources on the network.

\subsection{System Information and Predictions}
\label{System Information and Predictions}
The following is an explination the UITS system from which trace1 pulls it's packet information along with predicitions of how the data will look.

The UITS system consisnts of five Microsoft file server cluster nodes.  These blade servers are used to host home directories for all UConn users within the list of 88 departments.  These home directories are used to provide personal drive share space to facultiy, staff and students, along with at lest one small group of users.  Each server is capable of handling 1Gb of traffic in each direction (e.g. outbound and inbound traffic).  All together the five blade server system can in theory handle 10Gb of recieving and tranmitting data.  Some of these blade servers have local storage but the majority do not have any.  To the understanding of this paper, the blade servers are purposed purely for dealing with incoming traffic to the SAN storage that sits beihnd them.  This system does not currently implement load balancing, instead the servers are set up to spead the traffic load among four of the active cluster nodes while the fifth node is passive and purposed to take over in the case that any of the other nodes goes down. \\

The following are my predictions about what the data will tell me about the system.  First are the predictions based on what was learned from talking to people within UITS, after that are my general predictions.

From this paper's understanding of the file server system there are spikes of traffic that tend to happen during the night time.  The assumption is that the majority of this traffic will occur between 2am and 6am because this is when backups occur to the SAN system.  The point of note is that, however, it is not expected that we would see any of this traffic as the protocol used is not the SMB/CIFS protocol that is being analyzed by this paper.  The reasoning for this is that this traffic would be encrypted, therefore this traffic would appear as some other protocol.  Further more, any traffic that does occur during the duration of "day time hours" (i.e. 9am to 5pm) would be soley due to the actions taken by the users of this system (e.g facutly, staff, students). \\
Assumptions:
\begin{itemize}
\item Some backup traffic will be seen because traffic will be generated as the data being stored using this "oneline storage" is backed up to the SAN system.  Note: Any traffic past moving the data to the SAN system will \textbf{not} be seen.
\item All backup will be performed late night/early morning (e.g. 11pm-5am)
\item One general assumption is that these blade servers are "rock solid" and therefore should \textbf{not} ever go down.  If this is the case then the expectation is that we should see at most a transfer rate of 8Gb since the fifth server will not be in operation.  If we do find that there is a greater rate of transfer of data, then this means that the fifth server is actually helping with the traffic, not just acting as a backup in the case that any other blade server crashes or "goes down".
\end{itemize}

\subsection{Run Patterns}
\label{Run Patterns}

\subsection{Locating Performance Bottlenecks}
\label{Locating Performance Bottlenecks}

\section{Intuition Confirm/Change}
\label{Intuition Confirm/Change}

\subsection{Characterizations of Different Packet Types}
\label{Characterizations of Different Packet Types}

%\section{Related Work}
%\label{Related Work}
%
%\subsection{Anderson 2004 Paper}
%\label{Anderson 2004 Paper}
%This paper tackles the temporal inaccuracy of current day benchmarks \& the impact and errors produced due to these naive benchmarking tools.  Timing accuracy (issuing I/Os at the desired time) at high I/O rates is difficult to achieve on stock operating systems ~\cite{Anderson2004}.  Due to this inaccuracy, these may be introduction of substantial errors into observed system metrics when benchmarking I/O systems; including the use of these inaccurate tools for replaying traces or for producing synthetic workloads with known inter-arrival times ~\cite{Anderson2004}.  Anderson \textit{et al.} demonstrates the need for timing accuracy for I/O benchmarking in the context of replaying I/O traces.  Anderson \textit{et al.} showed that the error in perceived I/O response times can be as much as +350\% or -15\% by using naive benchmarking tools that have timing inaccuracies ~\cite{Anderson2004}.  Anderson \textit{et al.}'s measurements indicated that the accuracy achieved by using standard system calls is not adequate and that errors in issuing I/Os can lead to substantial errors in measurements of I/O statistics such as mean latency and number of outstanding I/Os.
%\begin{itemize}
%	\item 1. Timing in accuracy of benchmakrs can lead to error of +350\% or -15\% in perceived I/O response times.  Accuracy achieved using standard system calls \textbf{not} adequate and error in issuing I/Os leads to substantial I/O statistics errors
%	\item 2. "We currently lack tools to easily and accurately generate complex I/O workloads on modern storage systems".  \textit{\textbf{Result}}: May introduce substantial errors in observed system metrics when benchmark I/O systems use inaccurate tools
%	\item 3. I/O benchmarking widespread practice in storage industry and serves as basis for purchasing decisions, performance tuning studies and marketing campains.  "how does a given storage system perform for my workload?"  Benchmarking done by comparing I/O systems by subjecting them to known workloads
%	\item 4. Three general approaches based on trade-off between experimental complexity and resemblence to application
%	\begin{itemize}
%		\item 1. Connect system to production/test environment, run application, measure application metrics
%		\item 2. Collect traces from running application and replay them (after possible modification) back on test I/O system
%		\item 3. Generate sythetic workload and measure system performance
%	\end{itemize}
%	\item 5. Most studies assume issue accuracy using standard system calls adequate.  Measures indicate not the case and errorsin issuing I/O can lead to substantial errors in issuing I/O can lead to substantial errors in I/O statistic measurements (e.g. mean latency and number of outstanding I/Os
%	\item 6. Inaccuracies in scheduling I/Os may result in as much as a factor of 3.5 difference in measured response time and factor of 26 in measured queue sizes; Differences too large to ignore
%	\item 7. Timing accuracy and high through-put involves three challenges
%	\begin{itemize}
%		\item 1. Designing for peak performance requirements
%		\item 2. Coping with OS timing inaccuracy
%		\item 3. Working around unpredictable OS behavior
%		\begin{itemize}
%			\item 1. Standard OS mechanisms to keep time and issue I/Os; accuracy determined by scheduling granularity of underlying OS
%			\item 2. Accuracy of I/O scheduling contingent upon thread being scheduled at right time by OS scheduling boundaries \textit{or} flatten bursts
%		\end{itemize}
%		\item 4. Unpredictable performance effects due to interrupts; locking, resource contention, kernel scheduling intracacies
%		\item 5. Examples of performance effects
%		\begin{itemize}
%			\item 1. \textit{gettimeofday}() function (SMP) from multiple threads may cause locking to preserve clock invarience
%			\item 2. Thread moving from one CPU to another difficulty keeping track of wall clock time
%		\end{itemize}
%		\item 6. In higher load case the kernel gets more opportunities to schedule threads and hence more I/O issuing threads get scheduled at right time
%	\end{itemize}
%\end{itemize}
%
%\subsection{Ellard Ledlie 2003}
%\label{Ellard Ledlie 2003}
%This paper examines two workloads (research and email) to see if they resemble previously studied workloads, as well as performs several new analyses on the NFS protocol.  Trace-based analyses have guided and motivated contemporary file system design for the past two decades; where the original analysis of the 4.2BSD file system motivated many of the design decisions of the log-structured file system (LFS)~\cite{EllardLedlie2003}.  This paper also takes the stance that since the use of technology has expanded and evolved, this fundamental change in workloads needs to be traced to observe and understand the behavior.  "We believe that as the community of computer users has expanded and evolved there has been a fundamental change in the workloads seen by file servers, and that the research community must find ways to observe and measure these new workloads."~\cite{EllardLedlie2003}  Some of the contributions of this paper include new techniques for analyzing NFS traces along with tools to gather new anonymized NFS traces.  Leung \textit{et al.} (as well as Ellard \textit{et. al.}) also observed that much of the variance of load characterization statistics over time can be explained by high-level changes in the workload over time; despite, this correlation having been observed in many trace studies, its effects are usually ignored~\cite{EllardLedlie2003}.  The most noticeable change in their traces was the difference between peak and off-peak hours of operation.  This finding conveyed that time is a strong predictor of operation counts, amount of data transferred, and the read-write ratios for their CAMPUS (e.g. email) workload.
%
%\subsection{Ellard 2003}
%\label{Ellard 2003}
%This paper shows that the technology being actively researched gains improvement faster and that the technology that is not improved will end up being the bottleneck of the system.  Ellard and Seltzer give the example of how file system performance is steadily losing ground relative to CPU, memory, and even network performance.  Even though Ellard and Seltzer began their efforts to accurately measure the impact of changes to their system, they also discovered several other phenomena that interacted with the performance of the disk and file system in ways that had far more impact on the overall performance of the system than their improvements~\cite{Ellard2003}.  This paper loosely groups all benchmarks into two categories: micro benchmarks and macro/workload benchmarks, the difference between these two being that micro benchmarks measure specific low-level aspects of system performance while workload benchmarks estimate the performance of the system running a particular workload.
%
%\subsection{Leung 2008 Paper}
%\label{Leung 2008 Paper}
%Comparison of file access patterns (RO, WO, RW)
%\begin{itemize}
%	\item 1. Workloads more write-heavy than previously seen
%	\item 2. RW access patterns much more frequent
%	\item 3. Bytes transferred in much longer sequential runs
%	\item 4. Bytes transferred from much larger files
%	\item 5. Files live order of magnitude longer
%	\item 6. Most files not reopened once closed; If file re-opened, temporally related to previous closing of file
%\end{itemize}
%Files are infrequently shared by more than one client; over 76\% files never opened by more than one client.
%File sharing rarely concurrent and usually read-only; 5\% of files opened by multiple client are concurrent and 90\% of sharing is read only
%\textit{Compared to Previous Studies}
%\begin{itemize}
%	\item 1. Both of our workloads are more write-oriented. Read to write byte ratios have significantly decreased.
%	\item 2. Read-write access patterns have increased 30-fold relative to read-only and write-only access patterns.
%	\item 3. Most bytes are transferred in longer sequential runs. These runs are an order of magnitude larger.
%	\item 4. Most bytes transferred are from larger files. File sizes are up to an order of magnitude larger.
%	\item 5. Files live an order of magnitude longer. Fewer than 50\% are deleted within a day of creation.
%\end{itemize}
%\textit{New Observations}
%\begin{itemize}
%	\item 6. Files are rarely re-opened. Over 66\% are re-opened once and 95\% fewer than five times.
%	\item 7. Files re-opens are temporally related. Over 60\% of re-opens occur within a minute of the first.
%	\item 8. A small fraction of clients account for a large fraction of file activity. Fewer than 1\% of clients account for 50\% of file requests.
%	\item 9. Files are infrequently shared by more than one client. Over 76\% of files are never opened by more than one client.
%	\item 10. File sharing is rarely concurrent and sharing is usually read-only. Only 5\% of files opened by multiple clients are concurrent and 90\% of sharing is read-only.
%\end{itemize}
%\textit{List of interesting data points (comes from 'Table 3: Summary of trace statistics')}
%\begin{itemize}
%	\item Clients, Days, Data Read (GB), Data Written (GB), R:W I/O Ratio, R:W Byte Ratio, Total Operations
%	\item Operation Names: Session Create, Open, Close, Read, Write, Flush, Lock, Delete, File Stat, Set Attribute, Directory Read, Rename, Pipe Transactions
%\end{itemize}
%\textit{Table 4: Comparison of file access patterns - This figure gives good show of Read-Only, Write-Only \& Read-Write}
%\\\textit{Observations:}
%\begin{itemize}
%	\item 1) “Both of our workloads are more write-heavy than workloads studied previously”
%	\item 2) “Read-write access patterns are much more frequent compared to past studies”
%	\item 3) “Bytes are transferred in much longer sequential runs than in previous studies”
%	\item 4) Bytes are transferred from much larger files than in previous studies
%	\item 5) Files live an order of magnitude longer than in previous studies
%	\item 6) Most files are not re-opened once they are closed
%	\item 7) If a file is re-opened, it is temporally related to the previous close
%\end{itemize}
%
%\subsection{Orosz 2013 Paper}
%\label{Orosz 2013 Paper}
%\begin{itemize}
%	\item 1. Primary bottleneck in current timestamp resolution provided by Kernel is large deviation of deneration (timestamp generation) overhead decreases timestamp precision
%	\item 2. Simplifying the work of the kernel (i.e. time stamping process) will lead to lower packet loss
%	\item 3. "In network measurement, the precision of timestamping is a criterion more important than the low clock offset, especially for measuring packet inter-arrival times and round-trip delays at one single point of the network (e.g., active probing)"
%\end{itemize}
%
%\subsection{Dabir 2008 Paper}
%\label{Dabir 2008 Paper}
%\begin{itemize}
%	\item 1. "Since Dumpcap is a single threaded application, it was suspected that while it is busy writing to disk, because it is blocked by the I/O call, it is unable to handle newly arriving packets due to the small size of the kernel buffer which quickly fills up."
%	\item 2. Increasing amount of kernel level buffer associated with capture socket could lead to better capture speeds with Dumpcap
%\end{itemize}
%\textit{Note}: While (item 1) could be a source of packet loss, until this is tested on a trace system do not assume this is a key limiting factor.  It could be that by having Dumpcap write to RAM would alieviate this problem.  If this is the case, Dabir \& Matrawy attempted to overcome this limitation by having synchronization between two threads (using two semaphores) where \textbf{one} thread would store/buffer incoming packets and the \textbf{second} thread would write the packet information to disk
%
%\subsection{Narayan 2010 Paper}
%\label{Narayan 2010 Paper}
%\begin{itemize}
%	\item 1. Striping Data in parallel file system can bottleneck if file distribution parameters do not fit access patterns of applications
%	\item 2. Parallel application have five major models of I/O
%	\begin{itemize}
%		\item 1. Single output file shared by multiple nodes by ranges
%		\item 2. Large sequential read by single node at beginning of computation and large sequential write by single node at end of computation
%		\item 3. Checkpointing of state
%		\item 4. Metadata and read intensive - small data I/O, frequent directory lookups for reads
%		\item 5. Each node outputs to its own file
%	\end{itemize}
%	\item 3. Distributing I/O across more nodes not decrease IATs because files striped across all nodes which causes any Read or Write to access all nodes
%	\item 4. From Figure 5: As we see in the graphs and data provided, as the number of I/O nodes increases (as well as the number of processors/servers) the IATs decrease along with the number of I/O operations increase.  This leads to a larger \% of IATs occuring at lower times (e.g. < 1ms)
%	\item 5. From study on IATs, most parallel applications doing significant I/O increase the I/O frequency as the number of compute nodes increases.  \textbf{However}, scaling I/O nodes alone causes issue because increased I/O load is transferred to each I/O storage node
%	\item 6. "I/O access models that assume independence or randomness between requests are not valid"
%\end{itemize}
%
%\subsection{Skopko 2012 Paper}
%\label{Skopko 2012 Paper}
%\begin{itemize}
%	\item 1. Software based capture solutions heavily rely on OS's packet processing mechanism
%	\item 2. "Dumpcap itself uses relatively small system resources, however it is executed on a general purpose system that shares its resources between the running processes"
%	\item 3. Drivers typically operate in two different modes: interrupt mode and polling mode; importance of modes is dual
%	\begin{itemize}
%		\item 1. timestamps generated at enqueueing process reflect that tiem instead of moment of arrival at physical layer
%		\item 2. polling causes bursty packet forwarding, thus need for appropriate sized buffers to handle them
%	\end{itemize}
%	\item 4. Dumpcap has option called \textit{snaplength} to do truncation; compared to original measurement, smaller snaplength = fewer lost packets by Dumpcap
%\end{itemize}

\section{Conclusion}
\label{Conclusion}
\textit{Do the results show a continuation in the trend of traditional computer science workloads?}
On the outset of this work it was believed that the data collected and analyzed would follow similar behavior patterns seen in previous papers \textit{Cite?}.  Our initial results were confusing and most definitely did not meet expections.  One of these oddities was that during the day one would see a greater increase in writes instead of reads.  The frist assumption was that this is due to the system and how users interact with everything.
I belive that the greater number of writes comes from students doing intro work for different classes in which students are constantly saving their work while reading instructions from a single source.  One must also recall that this data itself has limited interpretation because only a small three week windows of infomration is being examined.  A better, and far more complete, image can be constructed using data captured from the following months, or more ideally, from an entire year's worth of data.  An other limitation of the results is the scope of the analysis is curbed and does not yet fully dissect all of the fields being passed in network communication.
The future work of this project would be to
\begin{itemize}
	\item 1. Complete the dissection analysis to include all captured fields from the originating pcap files.  
	\item 2. All DataSeries files (which are purposed for distribution) would be a single file per day's worth of communication; this may be possible with new additions to the DataSeries code but pcap limitations do not currently allow for this.  
	\item 3. Modulation of the capturing software would not only pull out information pertanent to the SMB/CIFS protocol, but would be able to pull multiple protocols which a user would be able to define prior to run-time.  
	\item 4. Better automation of the capturing system would remove the potential of human error cause loss of data.  Use of new DataSeries tools may allow for recovery of previously corrupted DataSeries files.
\end{itemize}

%references section
%\bibliographystyle{plain}
%\bibliography{body}

\begin{thebibliography}{99}

\bibitem{Leung2008} Andrew W.~Leung and Shankur Pasupathy and Garth Goodson and Ethan L.~Miller,
\emph{Measurement and Analysis of Large-Scale Network File System Workloads}, USENIX Annual Technical Conference (June 2008)

\bibitem{Ellard2003} Daniel Ellard and Margo Seltzer, \emph{NFS Tricks and Benchmarking Traps},
USENIX Annual Technical Conference (2003)

\bibitem{EllardLedlie2003} Daniel Ellard and Jonathan Ledlie and Pia Malkani and Margo Seltzer, \emph{
Passive NFS Tracking of Email and Research Workloads}, 2nd USENIX Conference on File and Storage Technologies (2003)

\bibitem{Anderson2004} Eric Anderson and Mahesh Kallahalla and Mustafa Uysal and Ram Swaiminnathan, \emph{
Buttress: A Toolkit for Flexible and High Fidelity I/O Benchmarking}, 3rd USENIX Conference of File and Storage Technologies (April 2004)

\bibitem{Orosz2013} P\'{e}ter Orosz and Tam\'{a}s Skopk\'{o}, \emph{
Multi-threaded Packet Timestamping for End-to-End QoS Evaluation}, The Eighth International Conference on Systems and Networks Communications (2013)

\bibitem{Dabir2008} Abes Dabir and Ashraf Matrawy, \emph{
Bottleneck Analysis of Traffic Monitoring using Wireshark}, IEEE (2008)

\bibitem{Narayan2010} Sumit Narayan and John A. Chandy, \emph{
I/O Characterization on a Parallel File System}, International Symposium on Performance Evaluation of Computer and Telecommunication Systems (2010)

\bibitem{Skopko2012} Tam\'{a}s Skopk\'{o}, \emph{
Loss Analysis of the Software-based Packet Capturing}, Carpathian Journal of Electronic and Computer Engineering 5 (2012)

\bibitem{MS-CIFS} \emph{Common Internet File System (CIFS) Protocol}, url{http://msdn.microsoft.com/en-us/library/ee442092.aspx}

\bibitem{MS-SMB} \emph{Server Message Block (SMB) Protocol}, url{http://msdn.microsoft.com/en-us/library/cc246231.aspx}

\bibitem{MS-SMB2} \emph{Server Message Block (SMB) Protocol Versions 2 and 3}, url{http://msdn.microsoft.com/en-us/library/cc246482.aspx}

\bibitem{Roselli2000} Drew Roselli and Jacob R. Lorch and Thomas E. Anderson, \emph{
A Comparison of File System Workloads}, Proceedings of 2000 USENIX Annual Technical Conference (June 2000)

\bibitem{Vogels1999} Werner Vogels, \emph{
File system usage in Windows NT 4.0}, Proceedings of the seventeenth ACM symposium on Operating systems principles (December 1999)

\bibitem{Meyer2012} Dutch T. Meyer and William J. Bolosky, \emph{
A Study of Practical Deduplication}, ACM Transactions on Storage (January 2012)

\bibitem{PFRING} \emph{PF\_RING High-speed packet capture, filtering and analysis}, url{http://www.ntop.org/products/pf\_ring/}

\bibitem{Traeger2008} Avishay Traeger and Erez Zadok and Nikolai Joukov and Charles P.~Wright, \emph{
A Nine Year Study of File System and Storage Benchmarking}, ACM Transactions on Storage (May 2008)

\bibitem{Kavalanekar2009} Swaroop Kavalanekar and Dushyanth Narayanan and Sriram Sankar and Eno Thereska and Kushagra Vaid and Bruce Worthington, \emph{
Measuring Database Performance in Online Services: A Trace-Based Approach}, Performance Evaluation and Benchmarking (2009)

\bibitem{Douceur1999} John R.~Douceur and William J.~Bolosky, \emph{
A Large-Scale Study of File-System Contents}, Proceedings of the 1999 ACM SIFMETRICS international conference on Measurement and modeling of computer systems (June 1999)

\bibitem{Ruemmler1993} Chris Ruemmler and John Wilkes, \emph{
UNIX disk access patterns}, Winter USENIX 1993 (January 1993)

\end{thebibliography}

\end{document}

% Revision History:
% designed specifically to meet requirements of
%  TCL97 committee.
% originally a template for producing IEEE-format articles using LaTeX.
%   written by Matthew Ward, CS Department, Worcester Polytechnic Institute.
% adapted by David Beazley for his excellent SWIG paper in Proceedings,
%   Tcl 96
% turned into a smartass generic template by De Clarke, with thanks to
%   both the above pioneers
% use at your own risk.  Complaints to /dev/null.
% make it two column with no page numbering, default is 10 point

% Munged by Fred Douglis <douglis@research.att.com> 10/97 to separate
% the .sty file from the LaTeX source template, so that people can
% more easily include the .sty file into an existing document.  Also
% changed to more closely follow the style guidelines as represented
% by the Word sample file.
% This version uses the latex2e styles, not the very ancient 2.09 stuff.
%

% Revised July--October 2002 by Bart Massey, Chuck Cranor, Erez
% Zadok and the FREENIX Track folks to ``be easier to use and work
% better''. Hah.  Major changes include transformation into a
% latex2e class file, better support for drafts, and some
% layout improvements.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% for Ispell:
% LocalWords:  workingdraft BCM ednote SubSections xfig SubSection joe